For models trained in sagemaker, I can just use invoke_endpoint() to hit endpoints.
For models trained with aws comprehend, I have to use classify_document() to hit endpoints.
In my code, I find myself having to use a different boto3.client for each service as well, which is messy.
To avoid my code getting more and more fragmented, is there any way that all my models can be managed and served via sagemaker? Even if the model is trained from a service outside of sagemaker?
Comprehend is an fully managed AI service and you will not have access to the trained models outside of comprehend. I believe you will not be able to deploy the comprehend model in sagemaker.